Currency of Value: Should Your AI Agent Charge for Time Saved, Revenue Gained, or Cost Cut?

Table of Contents

• The Three Currencies, and Why You Have to Pick One
• Time Saved: The Default Everyone Reaches For
• Cost Cut: The CFO's Favorite, and Your Ceiling
• Revenue Gained: The Highest Price and the Hardest Proof
• The Attribution Problem Nobody Wants to Talk About
• A Decision Framework: Matching Currency to Buyer and Workflow
• Why Your Currency and Your Pricing Mechanic Are Two Different Choices
• Insights Most People Overlook
• References

The Three Currencies, and Why You Have to Pick One

When founders talk about "value-based pricing" for agents, they're usually skipping a step. Before you decide how to charge, per task, per outcome, per resolution, you have to decide what value you're claiming to create in the first place. That claim is the currency. It's the thing you point to when a procurement officer asks, "Why is this worth the number on the contract?"

There are really only three answers that survive a budget review: you saved someone time, you cut a cost, or you brought in revenue. Marketing decks love to gesture at all three at once ("save 40 hours a week, slash operating costs, and grow your pipeline!"), but a pricing model can only have one primary anchor. The anchor is what the price scales against. If your price goes up when the customer saves more hours, you've picked time. If it goes up when their revenue grows, you've picked revenue. Trying to anchor on all three produces a pricing page nobody can model, which is its own kind of problem worth solving separately.

The choice matters because each currency routes you to a different buyer, a different proof burden, and a different ceiling on what you can charge. Get it wrong and you'll spend every renewal arguing about whether the value was real.

Time Saved: The Default Everyone Reaches For

Time saved is where almost every agent vendor starts, and it's easy to see why. It's intuitive, it's universal, and you can usually estimate it before the customer even signs. A support agent resolves a ticket in two minutes that used to take a human fifteen. A research agent compresses a four-hour competitive teardown into twenty minutes. The math practically writes itself.

The trap is that time saved is the weakest currency for justifying price, even though it's the strongest for generating a demo "wow." The reason is fungibility. When you tell a VP of Operations you saved her team 40 hours a week, her honest internal question is: and then what? If those 40 hours don't translate into a backfilled headcount, a canceled req, or measurable new output, the saving evaporates into "people are slightly less busy." Slightly-less-busy doesn't show up on any financial statement. It's real, but it's not bankable, and sophisticated buyers know it.

There's a deeper issue. Knowledge work rarely operates at full utilization, so the marginal hour you free up is often the least valuable hour, the slack time, not the bottleneck. Eliyahu Goldratt's theory of constraints made this point decades before AI agents existed: improvements anywhere other than the bottleneck are largely illusory. An agent that saves time on a non-constraint looks impressive in a usage dashboard and changes nothing about throughput.

Time-saved pricing works best in two specific situations. The first is when the saved time is immediately convertible, billable-hour professions (law, consulting, agencies) where an hour returned is an hour resold. The second is when you're pricing for an individual prosumer who literally has more work than hours, and any time back is pure relief. For most enterprise workflows, treat time saved as a top-of-funnel story, not a pricing anchor. It opens the conversation; it shouldn't set the number.

Cost Cut: The CFO's Favorite, and Your Ceiling

Cost cut is what time saved becomes once it grows up. Instead of "we save your team 40 hours," it's "we let you operate this function with three fewer FTEs," or "we reduce your per-ticket support cost from \$8 to \$1.20." Now you're speaking the only language procurement fully trusts: dollars that were going out the door and now aren't.

This is the currency behind the per-resolution support agents that have set the market's expectations, the Intercom Fin model and its many imitators, which a thoughtful breakdown of agent pricing economics from Andreessen Horowitz traces directly to the labor cost they displace. When you can point at a fully-loaded support rep costing \$60,000 a year and say your agent handles 50% of their volume, the cost-cut math is auditable, defensible, and survives the CFO's spreadsheet. That defensibility is worth a lot.

But cost cut has a structural ceiling that founders consistently underestimate: you can never charge more than the cost you eliminate, and in practice you can only capture a fraction of it. If you save a customer \$100, you might capture \$20 to \$40 of it, the rest is the buyer's incentive to do the deal at all. Worse, cost-cut framing invites a brutal comparison. The moment you anchor on "cheaper than a human," the buyer starts benchmarking you against the cheapest possible human: an offshore BPO, a contractor, the next vendor's quote. You've entered a race to the bottom that compresses margins over time, especially as inference costs fall and competitors pass those savings through.

Cost cut is the right currency when your agent genuinely replaces a well-understood, well-priced line item, support, data entry, document processing, basic bookkeeping. It's the wrong currency when the work was never a clean cost center to begin with, because then you're inventing a baseline the buyer can dispute at every renewal.

Revenue Gained: The Highest Price and the Hardest Proof

Revenue gained is the currency everyone wants and few can credibly claim. If your agent books meetings that turn into deals, recovers abandoned carts, upsells existing accounts, or generates qualified pipeline, you're no longer a cost on the P&L, you're a growth lever. And growth levers command a completely different price. McKinsey's work on generative AI's economic potential found that the largest value pools sit not in cost reduction but in revenue-generating functions like sales and marketing, precisely because the upside is uncapped in a way cost savings never are.

Here's the asymmetry that makes revenue so attractive: there's no ceiling. A cost-cut agent that saves \$100K caps your revenue at some slice of \$100K. A revenue agent that helps close \$2M in new business can justify a six-figure fee without the buyer blinking, because they're sharing upside, not surrendering budget. This is why the most aggressive value-based and success-fee pricing experiments cluster in revenue-adjacent agents.

The catch, and it's a big one, is attribution. Cost savings are subtractive and clean: the headcount line went down, full stop. Revenue is additive and contested: did the agent cause the deal, or did it touch a deal that would have closed anyway? Every CMO has lived through the multi-touch attribution wars, and they will apply that same skepticism to your agent. If you can't draw a defensible line from agent action to closed revenue, your beautiful revenue-share pricing collapses into a quarterly argument.

Revenue gained works when the causal chain is short and observable: an agent that autonomously recovers a churning customer, completes a checkout, or sources a lead that converts within a tight window. It fails when your agent is one of fifteen touches in a six-month enterprise sales cycle. The length and cleanliness of the attribution chain, not the size of the prize, determines whether revenue is a currency you can actually bill against.

The Attribution Problem Nobody Wants to Talk About

Notice the pattern across all three currencies: the price you can charge runs opposite to the ease of proving it. Time saved is trivial to measure and worth the least. Revenue gained is worth the most and nearly impossible to attribute cleanly. Cost cut sits in the productive middle, provable enough to defend, valuable enough to matter.

This inverse relationship is the single most important thing to internalize about value currencies, and most pricing pages ignore it entirely. The vendors who win don't pick the currency with the highest theoretical value. They pick the highest-value currency they can prove with a measurement the buyer already trusts. Proof beats prize.

Practically, this means your currency choice is downstream of your instrumentation. Can you measure a clean baseline? Can you isolate your agent's contribution from confounding factors? Can the customer audit your numbers without taking your word for it? If the answer to any of these is no, you should step down one rung, from revenue to cost, or from cost to time, until you reach a currency you can actually substantiate. A defensible cost-cut number beats an unprovable revenue claim every single time, because the unprovable claim dies at renewal when the champion who believed you has moved on and their replacement asks for evidence you can't produce. This is also why defining and auditing the outcome itself is becoming the real battleground in agent pricing.

A Decision Framework: Matching Currency to Buyer and Workflow

Strip away the theory and the choice comes down to four questions. Run your agent through them in order.

1. Whose budget are you reaching for?

The currency must match the buyer's mandate. A line-of-business operations leader is measured on efficiency and cost, speak cost cut. A revenue leader (CRO, VP Sales, CMO) is measured on growth, speak revenue gained, if you can prove it. An individual practitioner or small team buying with a credit card cares about their own throughput, time saved lands fine there. Pitching a revenue story to a cost-center owner, or a time-savings story to a CRO, is a currency mismatch that stalls deals no matter how good the agent is.

2. How clean is the attribution chain?

Count the steps and the confounders between your agent's action and the value. One step, no confounders (agent resolves ticket → cost avoided) supports cost-cut pricing. Many steps, many confounders (agent contributes to a deal among fifteen touches) does not support revenue pricing, fall back to cost or activity. Be ruthlessly honest here; this is where optimism kills pricing models.

3. Is there a trusted baseline?

You can only charge against a saving or gain if both sides agree on the "before." Replacing a known \$8-per-ticket cost gives you a baseline. "Growing pipeline" without a controlled comparison does not. No agreed baseline means no defensible value claim, which pushes you toward simpler activity-based or per-task pricing until you can establish one.

4. How much upside are you leaving on the table?

If your agent genuinely drives revenue and you can prove it, anchoring on cost cut is malpractice, you're capping yourself at a fraction of a number far smaller than the value you create. The discipline is to climb as high up the currency ladder (time → cost → revenue) as your proof will support, and not one rung higher.

The honest takeaway: most agents today should anchor on cost cut, because it's the currency where value and provability overlap best. Reserve revenue framing for the narrow set of agents with short, observable causal chains, and use time saved as a marketing hook rather than a billing anchor.

Why Your Currency and Your Pricing Mechanic Are Two Different Choices

One more distinction founders blur constantly. The currency (what value you claim) and the mechanic (how you structure the charge) are separate decisions, and you can mix them.

You can anchor on cost cut and still bill per seat, per task, or per outcome. A support agent priced per resolution uses an outcome mechanic with a cost-cut currency, the resolution is the unit, the displaced agent cost is the justification. You could equally take a cost-cut currency and wrap it in a flat subscription that's simply benchmarked against the FTEs it replaces. The currency lives in your sales narrative and your value justification; the mechanic lives on the order form.

This separation is liberating, because it means a hard-to-meter currency doesn't force a hard-to-meter mechanic. Revenue is messy to attribute in real time, but you can still anchor on a revenue narrative while billing a clean monthly platform fee with a usage component on top. The buyer hears "this grows your revenue"; the contract says "\$X per month plus \$Y per qualified lead." Keeping the two layers distinct lets you tell the highest-value story your proof supports while billing on the cleanest mechanic your operations can handle, which is exactly the balance well-designed hybrid models are reaching for.

Insights Most People Overlook

The most provable currency usually beats the most valuable one, and founders pick backwards. Vendors instinctively reach for the biggest number (revenue!) because it sounds best in a pitch. But pricing isn't won in the pitch; it's defended at renewal, where unprovable value claims go to die. The vendor who anchors on a boring, auditable cost-cut number out-survives the one chasing an unattributable revenue share, even though the revenue story looked better on the deck.

Time saved is a demo currency masquerading as a pricing currency. It exists to create the "holy cow" moment in the sales cycle, then it should hand off. Vendors who actually bill on hours-saved are quietly training buyers to interrogate utilization, "but were those hours real, did anyone get freed up?", which is the one question that deflates the whole deal. Use the hours to open; never let them set the price.

Cost-cut framing secretly hands your pricing power to the labor market. The moment you say "cheaper than a human," you've pegged your ceiling to the cheapest available human, and as offshore labor, contractors, and rival agents push that floor down, your price erodes with it. The escape hatch is to migrate the narrative from "cheaper than the person doing this" to "does work no person was doing at all," which detaches your price from the wage benchmark entirely.

The currency you choose silently selects your competitor. Anchor on cost and you're benchmarked against BPOs and headcount budgets. Anchor on revenue and you're benchmarked against marketing spend and agency retainers, a far more generous comparison set. Two agents doing nearly identical work can command wildly different prices purely because one framed its value against a cheap baseline and the other against an expensive one. The framing isn't spin; it's the most consequential pricing decision you'll make.

As inference costs collapse, cost-cut pricing becomes a trap and revenue pricing becomes the moat. When the underlying model cost drops 10x, every cost-anchored vendor faces customer pressure to pass the savings through, your price is mechanically linked to a falling input. Revenue-anchored pricing is insulated: the buyer doesn't care what your inference costs, only what you grew. The vendors who can credibly make the revenue claim will widen their margins precisely as the cost-anchored field gets commoditized.

References

• The Economics of AI Agents, Andreessen Horowitz (a16z)
• The Economic Potential of Generative AI: The Next Productivity Frontier, McKinsey & Company